Automated vs manual accessibility testing
Learn what automated accessibility checks detect, what requires human review, and how agencies can combine both in a practical workflow.
Automated and manual accessibility testing are complementary layers. Automation is repeatable, fast, and well suited to code-detectable rules. Manual testing brings context: whether an interaction is understandable, operable, and coherent for a person using different input or assistive technology.
What automated testing does well
Automated tools can find issues such as missing accessible names, invalid ARIA relationships, certain contrast failures, duplicate identifiers, and structural problems that can be expressed as deterministic rules. They are valuable during development, at release, and in scheduled accessibility regression monitoring because the same checks can be repeated consistently.
Automation is especially useful for preventing known issue classes from silently returning. It can also create a prioritized starting point for investigation. A passing automated scan, however, does not mean a page is accessible or conforms to a standard.
What needs human review
People must assess whether alternative text communicates the right meaning, keyboard focus follows a sensible path, error recovery is understandable, headings describe the content, controls behave predictably, and zoom or reflow preserves the task. Assistive-technology testing can reveal interaction problems that cannot be inferred reliably from static rules.
Manual review should include realistic journeys rather than disconnected component checks. Complete a form, navigate a menu, recover from an error, and move through a conversion path using the relevant inputs and output modes.
| Question | Automated | Manual |
|---|---|---|
| Does an image have an alt attribute? | Yes | Yes |
| Is the alt text appropriate in context? | No | Yes |
| Does a control expose a name? | Often | Yes |
| Is the keyboard order logical? | Limited | Yes |
| Did a known detectable issue reappear? | Yes | Optional verification |
| Is the overall task understandable? | No | Yes |
Combine the layers in an agency workflow
Run automation continuously on representative pages. Review new findings after confirmed changes rather than repeatedly triaging the full historic backlog. Schedule manual testing around launches, major component changes, critical journeys, and an agreed recurring cadence. Record scope and limitations in client communication.
When a manual review identifies a failure that can be represented by an automated assertion, add that protection to development or monitoring. When it cannot, keep it in a human checklist with a clear owner and schedule. White-label accessibility reports can present both layers without merging their conclusions.
This division is not a percentage-based shortcut for deciding whether a site is “done.” It is a way to assign each kind of question to the method capable of answering it.